A Modular Flask-Based Platform for Real-Time IoT Device Classification and Traffic Analysis

Abstract

The rapid expansion of connected smart devices has introduced significant challenges in managing and securing complex IoT ecosystems, particularly due to the diversity of device types and their dynamic communication patterns. This work proposes an intelligent IoT device identification framework that classifies networked entities by analyzing their traffic behavior rather than relying on static signatures. The system extracts key statistical and temporal features such as transmission intervals, packet size distributions, and session durations to construct distinctive behavioral profiles for each device category. To mitigate the issue of skewed datasets—where frequently occurring devices dominate learning—the approach integrates Adaptive Synthetic (ADASYN) sampling to enhance minority class representation and improve generalization. The classification engine is built upon a comparative analysis of multiple supervised learning techniques, including Gaussian Naive Bayes (GNB), Multinomial Naive Bayes (MNB), Decision Tree Classifier (DTC), and a novel Greedy Tree Classifier (GTC), which emphasizes interpretable rule-based decision structures while maintaining strong predictive capability. The framework is deployed through a lightweight Flaskbased web interface that supports both data visualization for exploratory insights and real-time device prediction. Performance evaluation demonstrates that the GTC model consistently delivers improved classification effectiveness, particularly in distinguishing heterogeneous IoT devices, thereby enabling scalable, automated, and secure network management solutions.